19
Immediate commands
Contents
19.1
19.2
19.3
19.1
Overview
19.1.1 Examples
19.1.2 A list of the immediate commands
The display command
The power command
Overview
An immediate command is a command that obtains data not from the data stored in memory but
from numbers typed as arguments. Immediate commands, in effect, turn Stata into a glorified hand
calculator.
There are many instances when you may not have the data, but you do know something about the
data, and what you know is adequate to perform statistical tests. For instance, you do not have to
have individual-level data to obtain the standard error of the mean, and thereby a confidence interval,
if you know the mean, standard deviation, and number of observations. In other instances, you may
actually have the data, and you could enter the data and perform the test, but it would be easier if
you could just ask for the statistic based on a summary. For instance, you flip a coin 10 times, and it
comes up heads twice. You could enter a 10-observation dataset with two ones (standing for heads)
and eight zeros (meaning tails).
Immediate commands are meant to solve those problems. Immediate commands have the following
properties:
1. They never disturb the data in memory. You can perform an immediate calculation as an aside
without changing your data.
2. The syntax for these commands is the same, the command name followed by numbers, which
are the summary statistics from which the statistic is calculated. The numbers are almost always
summary statistics, and the order in which they are specified is in some sense “natural”.
3. Immediate commands all end in the letter i , although the converse is not true. Usually, if there
is an immediate command, there is a nonimmediate form also, that is, a form that works on the
data in memory. For every statistical command in Stata, we have included an immediate form if
it is reasonable to assume that you might know the requisite summary statistics without having
the underlying data and if typing those statistics is not absurdly burdensome.
4. Immediate commands are documented along with their nonimmediate counterparts. Thus, if you
want to obtain a confidence interval, whether it be from summary data with an immediate command
or using the data in memory, use the table of contents or index to discover that [R] ci discusses
confidence intervals. There, you learn that ci calculates confidence intervals by using the data in
memory and that cii does the same with the data specified immediately following the command.
1
2
19.1.1
[ U ] 19 Immediate commands
Examples
Example 1
Let’s take the example of confidence intervals. Professional papers often publish the mean, standard
deviation, and number of observations for variables used in the analysis. Those statistics are sufficient
for calculating a confidence interval. If we know that the mean mileage rating of cars in some sample
is 24, that the standard deviation is 6, and that there are 97 cars in the sample, we can calculate
. cii 97 24 6
Variable
Obs
Mean
97
24
Std. Err.
[95% Conf. Interval]
.6092077
22.79073
25.20927
We learn that the mean’s standard error is 0.61 and its 95% confidence interval is [ 22.8, 25.2 ].
To obtain this, we typed cii (the immediate form of the ci command) followed by the number
of observations, the mean, and the standard deviation. We knew the order in which to specify the
numbers because we had read [R] ci.
We could use the immediate form of the ttest command to test the hypothesis that the true mean
is 22:
. ttesti 97 24 6 22
One-sample t test
x
Obs
Mean
97
24
mean = mean(x)
Ho: mean = 22
Ha: mean < 22
Pr(T < t) = 0.9993
Std. Err.
Std. Dev.
.6092077
6
[95% Conf. Interval]
22.79073
25.20927
t =
3.2830
degrees of freedom =
96
Ha: mean != 22
Ha: mean > 22
Pr(|T| > |t|) = 0.0014
Pr(T > t) = 0.0007
The first three numbers were as we specified in the cii command. ttesti requires a fourth number,
which is the constant against which the mean is being tested; see [R] ttest.
Example 2
We mentioned flipping a coin 10 times and having it come up heads twice. The 99% confidence
interval can also be obtained from ci:
. cii 10 2, level(99)
Variable
Obs
Mean
10
.2
Std. Err.
Binomial Exact
[99% Conf. Interval]
.1264911
.0108505
.6482012
In the previous example, we specified cii with three numbers following it; in this example, we
specify 2. Immediate commands often determine what to do by the number of arguments following
the command. With two arguments, ci assumes that we are specifying the number of trials and
successes from a binomial experiment; see [R] ci.
[ U ] 19 Immediate commands
3
The immediate form of the bitest command performs exact hypothesis testing:
. bitesti 10 2 .5
N
Observed k
Expected k
10
2
5
Pr(k >= 2)
= 0.989258
Pr(k <= 2)
= 0.054688
Pr(k <= 2 or k >= 8) = 0.109375
Assumed p
Observed p
0.50000
(one-sided test)
(one-sided test)
(two-sided test)
0.20000
For a full explanation of this output, see [R] bitest.
Example 3
Stata’s tabulate command makes tables and calculates various measures of association. The
immediate form, tabi, does the same, but we specify the contents of the table following the
command:
. tabi 5 10 \ 2 14
row
1
2
Total
col
1
5
2
7
Fisher’s exact =
1-sided Fisher’s exact =
2
Total
10
14
15
16
24
31
0.220
0.170
The tabi command is slightly different from most immediate commands because it uses ‘\’ to
indicate where one row ends and another begins.
4
[ U ] 19 Immediate commands
19.1.2
A list of the immediate commands
Command
Reference
Description
bitesti
[R] bitest
Binomial probability test
cci
csi
iri
mcci
[ST] epitab
Tables for epidemiologists
cii
[R] ci
Confidence intervals for means, proportions, and counts
esizei
[R] esize
Effect size based on mean comparison
prtesti
[R] prtest
Tests of proportions
sdtesti
[R] sdtest
Variance comparison tests
symmi
[R] symmetry
Symmetry and marginal homogeneity tests
tabi
[R] tabulate twoway
Two-way tables of frequencies
ttesti
[R] ttest
Mean comparison tests
twoway pci
twoway pcarrowi
twoway scatteri
[G-2] graph twoway pci
[G-2] graph twoway pcarrowi
[G-2] graph twoway scatteri
Paired-coordinate plot with spikes or lines
Paired-coordinate plot with arrows
Twoway scatterplot
19.2
The display command
display is not really an immediate command, but it can be used as a hand calculator.
. display 2+5
7
. display sqrt(2+sqrt(3^2-4*2*-2))/(2*3)
.44095855
See [R] display.
19.3
The power command
power is not technically an immediate command because it does not do something on typed
numbers that another command does on the dataset. It does, however, work strictly on numbers you
type on the command line and does not disturb the data in memory.
power performs power and sample-size analysis. See Stata Power and Sample-Size Reference
Manual.